Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A system for generating capsule neural networks for enhancing image processing platforms, the system comprising: at least one transitory storage device; and at least one processing device coupled to the at least one non-transitory storage device, wherein the at least one processing device is configured to: generate capsule neural network based on instructions received form at least one user; transfer learning from an existing image processing platform to train the capsule neural network; receive input from one or more devices and provide the input to the existing image processing platform comprising a convolutional neural network, wherein the convolutional neural network processes the input; activate the capsule neural network, wherein upon activation the capsule neural network: validates processing of the convolutional neural network; and identifies that the validation of the processing of the convolutional neural network is unsuccessful; in response to the unsuccessful validation, extract a part of the input that is associated with the unsuccessful validation; and transfer the part of the input to the capsule neural network to retrain the capsule neural network.
The system enhances image processing platforms by integrating capsule neural networks (CapsNet) to improve accuracy and robustness. Traditional convolutional neural networks (CNNs) excel at feature extraction but struggle with spatial hierarchies and viewpoint invariance, leading to errors in complex image recognition tasks. This system addresses these limitations by combining CNNs with CapsNets, which better capture spatial relationships and hierarchical features. The system includes a processing device and storage, configured to generate a CapsNet based on user instructions. It employs transfer learning to train the CapsNet using pre-existing CNN models, leveraging their learned features while mitigating their weaknesses. During operation, the system receives input from devices and processes it through the CNN. The CapsNet then validates the CNN's output. If validation fails, the system isolates the problematic input segment and retrains the CapsNet using this data, refining its performance over time. This adaptive approach ensures continuous improvement in image processing accuracy, particularly for challenging scenarios where CNNs alone underperform. The system thus enhances reliability in applications like medical imaging, autonomous vehicles, and quality control, where precise image interpretation is critical.
2. The system of claim 1 , wherein in response to retraining the capsule neural network, the at least one processing device is further configured to: determine accuracy associated with processing of the capsule neural network; identify that the accuracy associated with the processing of the capsule neural network is greater than predetermined threshold value; and in response to determining that the accuracy associated with the processing of the capsule neural network is greater than the predetermined threshold value, decommission the convolutional neural network and replace the convolutional neural network with the capsule neural network.
The system relates to machine learning, specifically the transition from convolutional neural networks (CNNs) to capsule neural networks (CapsNets) for improved accuracy. CNNs are widely used but struggle with spatial hierarchies and viewpoint invariance, leading to suboptimal performance in tasks requiring fine-grained feature recognition. CapsNets address these limitations by preserving spatial relationships and hierarchical features through dynamic routing and vector-based representations. The system monitors the accuracy of a CapsNet during retraining. If the CapsNet's accuracy exceeds a predefined threshold, the system automatically decommissions the existing CNN and replaces it with the CapsNet. This ensures that the system leverages the superior performance of CapsNets when they demonstrate sufficient accuracy, improving overall model efficiency and reliability. The transition is automated, reducing manual intervention and ensuring seamless integration of the more advanced neural network architecture. This approach is particularly useful in applications where accuracy and adaptability are critical, such as image recognition, natural language processing, and autonomous systems.
3. The system of claim 2 , wherein the at least one processing is configured to determine the accuracy based on comparing processing of the capsule neural network with one or more metrics.
A system for evaluating the performance of a capsule neural network is disclosed. The system addresses the challenge of assessing the accuracy and reliability of capsule neural networks, which are neural network architectures designed to better model hierarchical relationships between features in data. The system includes a processing unit configured to analyze the capsule neural network's output by comparing it against one or more predefined metrics. These metrics may include statistical measures, error rates, or other performance indicators relevant to the specific application of the neural network. The processing unit evaluates the network's accuracy by determining how closely the network's predictions align with the expected or ground truth values, as measured by the selected metrics. The system may also include additional components, such as input interfaces for receiving data and output interfaces for displaying or transmitting the accuracy results. The overall goal is to provide a robust method for validating the performance of capsule neural networks in various applications, ensuring their reliability in tasks such as image recognition, natural language processing, or other machine learning applications.
4. The system of claim 1 , wherein in response to identification of the unsuccessful validation, the at least one processing device is further configured to transfer a notification to at least one computing system.
5. The system of claim 1 , wherein the capsule neural network in response to identifying that the validation of the processing of the convolutional neural network is unsuccessful, automatically performs self-retraining based on the part of the input that is associated with the unsuccessful validation; automatically identifies that accuracy associated with processing of inputs is greater than a predetermined threshold; and automatically decommissions the convolutional neural network.
6. The system of claim 1 , wherein the input received from the one or more devices comprises limited data.
A system for processing input data from one or more devices, where the input data is limited in scope or quantity. The system includes a data processing module that receives and analyzes the limited input data, which may include partial, incomplete, or restricted information. The system further includes a data augmentation module that supplements the limited input data with additional information from external sources or internal databases to enhance its utility. A decision-making module then evaluates the augmented data to generate actionable insights or recommendations. The system may also include a feedback mechanism to refine the data processing and augmentation steps based on user interactions or system performance. The limited input data may originate from sensors, user inputs, or other data sources where full or complete data is not available or practical to obtain. The system is designed to handle scenarios where data is constrained by technical, privacy, or operational limitations, ensuring reliable decision-making despite incomplete information. The overall goal is to maximize the value of limited data inputs through intelligent augmentation and analysis.
7. The system of claim 1 , wherein the capsule neural network comprises one or more capsules, wherein each of the one or more capsules process a distinct feature associated with the input.
8. A computer program product for generating capsule neural networks for enhancing image processing platforms, the computer program product comprising a non-transitory computer-readable storage medium having computer executable instructions for causing a computer processor to perform the steps of: generating capsule neural network based on instructions received form at least one user; transferring learning from an existing image processing platform to train the capsule neural network; receiving input from one or more devices and provide the input received from one or more devices and provide the input to the existing image processing platform comprising a convolutional neural network, wherein the convolutional neural network processes the input; activating the capsule neural network, wherein upon activation the capsule neural network: validates processing of the convolutional neural network; and identifies that the validation of the processing of the convolutional neural network is unsuccessful; in response to the unsuccessful validation, extracting a part of the input that is associated with the unsuccessful validation; and transferring the part of the input to the capsule neural network to retrain the capsule neural network.
This invention relates to enhancing image processing platforms using capsule neural networks (CapsNets) to improve accuracy and reliability. Traditional convolutional neural networks (CNNs) often struggle with spatial hierarchies and viewpoint invariance, leading to errors in image recognition tasks. The invention addresses this by integrating a CapsNet with an existing CNN-based image processing system. The CapsNet is trained using transfer learning from the CNN, leveraging pre-existing knowledge to improve performance. The system receives input from one or more devices, which the CNN processes initially. The CapsNet then validates the CNN's output. If the validation fails, the system identifies the problematic input segment and retrains the CapsNet using that segment. This iterative process enhances the CapsNet's ability to handle complex image data, reducing errors and improving overall system accuracy. The invention automates the integration of CapsNets into existing image processing workflows, providing a scalable solution for improving image recognition performance without requiring a complete system overhaul.
9. The computer program product of claim 8 , wherein in response to retraining the capsule neural network, the computer executable instructions further cause the computer processor to: determine accuracy associated with processing of the capsule neural network; identify that the accuracy associated with the processing of the capsule neural network is greater than predetermined threshold value; and in response to determining that the accuracy associated with the processing of the capsule neural network is greater than the predetermined threshold value, decommission the convolutional neural network and replace the convolutional neural network with the capsule neural network.
10. The computer program product of claim 9 , wherein the computer executable instructions cause the computer processor to determine the accuracy based on comparing processing of the capsule neural network with one or more metrics.
A capsule neural network system evaluates the accuracy of its processing by comparing performance against predefined metrics. The system includes a neural network with capsules that encode spatial hierarchies of features, where each capsule outputs a vector representing both probability and pose information. The system processes input data through these capsules, generating outputs that are compared against reference metrics to assess accuracy. These metrics may include classification accuracy, pose estimation precision, or other performance indicators. The system dynamically adjusts processing based on the comparison results to improve performance. This approach enhances the reliability of capsule networks by ensuring outputs meet specified accuracy thresholds, addressing challenges in traditional neural networks where spatial relationships between features are often lost. The system is particularly useful in applications requiring precise feature extraction and pose estimation, such as image recognition, robotics, and autonomous systems. By integrating metric-based evaluation, the system provides a robust framework for validating and refining neural network outputs.
11. The computer program product of claim 8 , wherein in response to identification of the unsuccessful validation, the computer executable instructions further cause the computer processor to transfer a notification to at least one computing system.
12. The computer program product of claim 8 , wherein the capsule neural network comprises one or more capsules, wherein each of the one or more capsules process a distinct feature associated with the input.
13. The computer program product of claim 8 , wherein the capsule neural network in response to identifying that the validation of the processing of the convolutional neural network is unsuccessful, automatically performs self-retraining based on the part of the input that is associated with the unsuccessful validation; automatically identifies that accuracy associated with processing of inputs is greater than a predetermined threshold; and automatically decommissions the convolutional neural network.
This invention relates to a computer program product for improving the reliability of neural network processing, particularly in systems where a convolutional neural network (CNN) is used for input processing. The problem addressed is the potential for CNNs to produce inaccurate or unreliable outputs, which can lead to system failures or degraded performance. The solution involves a capsule neural network (CapsNet) that monitors the CNN's performance and takes corrective actions when validation fails. The CapsNet evaluates the CNN's processing of input data and determines whether validation is successful. If validation fails, the CapsNet automatically retrains itself using the portion of the input data associated with the unsuccessful validation. This self-retraining process allows the CapsNet to refine its processing capabilities. Once retraining is complete, the CapsNet assesses the accuracy of its processing. If the accuracy exceeds a predetermined threshold, the CapsNet automatically decommissions the CNN, ensuring that only reliable processing continues. This approach enhances system robustness by dynamically adjusting to performance issues without manual intervention. The invention is particularly useful in applications requiring high reliability, such as autonomous systems, medical diagnostics, or real-time decision-making environments.
14. The computer program product of claim 8 , wherein the input received from the one or more devices comprises limited data.
15. A computerized method for generating capsule neural networks for enhancing image processing platforms, the method comprising: generating capsule neural network based on instructions received form at least one user; transferring learning from an existing image processing platform to train the capsule neural network; receiving input from one or more devices and provide the input to the existing image processing platform comprising a convolutional neural network, wherein the convolutional neural network processes the input; activating the capsule neural network, wherein upon activation the capsule neural network: validates processing of the convolutional neural network; and identifies that the validation of the processing of the convolutional neural network is unsuccessful in response to the unsuccessful validation, extracting a part of the input that is associated with the unsuccessful validation; and transferring the part of the input to the capsule neural network to retrain the capsule neural network.
16. The computerized method of claim 15 , wherein retraining the capsule neural network further comprises: determining accuracy associated with processing of the capsule neural network; identifying that the accuracy associated with the processing of the capsule neural network is greater than predetermined threshold value; and in response to determining that the accuracy associated with the processing of the capsule neural network is greater than the predetermined threshold value, decommissioning the convolutional neural network and replace the convolutional neural network with the capsule neural network.
This invention relates to machine learning systems that improve performance by dynamically replacing convolutional neural networks (CNNs) with capsule neural networks (CapsNets) under specific conditions. The problem addressed is the static nature of neural network architectures, which may not adapt optimally to changing data distributions or performance requirements. The solution involves a computerized method for retraining a CapsNet and conditionally replacing a CNN with it based on accuracy metrics. The method monitors the accuracy of a CapsNet during processing. If the accuracy exceeds a predetermined threshold, the system decommissions the CNN and replaces it with the CapsNet. This replacement ensures that the system leverages the more advanced CapsNet architecture when it demonstrates superior performance. The approach improves system adaptability and efficiency by dynamically selecting the optimal neural network architecture based on real-time performance evaluation. The invention is particularly useful in applications requiring high accuracy and adaptability, such as image recognition, natural language processing, and other domains where neural network performance is critical.
17. The computerized method of claim 16 , wherein determining the accuracy is based on comparing processing of the capsule neural network with one or more metrics.
18. The computerized method of claim 15 , wherein in response to identification of the unsuccessful validation, the computer executable instructions further cause the computer processor to transfer a notification to at least one computing system.
This invention relates to computerized methods for validating data or processes, particularly in systems where validation failures require automated notifications. The method involves a computer processor executing instructions to validate data or operations against predefined criteria. If validation fails, the system automatically generates and transfers a notification to at least one computing system, ensuring timely awareness of issues. The notification may include details about the failure, such as the type of validation error, the data or process involved, and any relevant context. This automated alerting mechanism helps maintain system integrity, reduce manual oversight, and enable rapid corrective action. The method may be applied in various domains, including financial transactions, data processing, or system monitoring, where validation failures could lead to errors or security risks. The notification can be sent via network protocols, APIs, or other communication channels to ensure recipients are promptly informed. This approach enhances reliability and responsiveness in automated validation workflows.
19. The computerized method of claim 15 , wherein the capsule neural network comprises one or more capsules, wherein each of the one or more capsules process a distinct feature associated with the input.
A computerized method involves using a capsule neural network to process input data, where the network includes multiple capsules, each dedicated to analyzing a distinct feature of the input. The capsules are designed to capture and represent specific characteristics of the input data, such as spatial hierarchies or relationships between features, in a way that preserves their relative positional and structural information. This approach improves the network's ability to recognize and interpret complex patterns, particularly in tasks like image or object recognition, where understanding the spatial relationships between features is critical. The method enhances traditional neural networks by incorporating dynamic routing mechanisms that allow capsules to communicate and adjust their outputs based on the relevance of their features to the overall input. This ensures that the network can adaptively focus on the most informative features while suppressing irrelevant or noisy data. The use of capsules helps mitigate common issues in deep learning, such as the loss of spatial information in convolutional layers, leading to more accurate and robust predictions. The method is particularly useful in applications requiring high precision, such as medical imaging, autonomous systems, and advanced pattern recognition.
20. The computerized method of claim 15 , wherein determining the accuracy based on comparing processing of the capsule neural network with one or more metrics.
Unknown
April 13, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.